Leverage map: where this fits in the Galaxy ecosystem

TL;DR. This project is a per-tool engine. Its value multiplies when it plugs into the places Galaxy tools are written, reviewed, published, and catalogued. This page maps those points, honestly tagged Exists / Potential / Roadmap, and lists the source repositories the repo-explainer skill draws on for examples and evidence. Nothing here claims an integration that has been built unless it says Exists.

Integration points

Where the format / upgrade / check engine could deliver value, and how mature that connection is today:

Where	What it would do	Maturity
AI agents (MCP)	Agents call `format`/`upgrade`/`check`/`list_*` directly while authoring or reviewing tools.	Exists: the MCP server ships (vision Goal 1).
`galaxy-mcp` (sibling MCP)	Coexist in the agent ecosystem; complementary surface.	Exists (separate project): positioning, not a built link.
`galaxy-language-server` (editor LSP)	Surface fixes as editor quick-fixes, reaching authors at write-time, before a PR exists.	🟡 In progress: the Tier-B rename API (`rename_param_plan`) shipped and `galaxy-tool-source` is on PyPI; a “Rename Symbol” + “Find References” binding is an open PR (galaxyproject/galaxy-language-server#331, CI-green against the PyPI release, cross-file across imported macros), initial maintainer review addressed, awaiting follow-up. A second binding, version-tokenization Code Actions over `tokenize_version_plan` (define tokens inline, or extract to a new `macros.xml` via a `WorkspaceEdit` + `CreateFile`), is built and validated against the published 0.2.0, awaiting its upstream PR.
`planemo` / `planemo-ci-action`	Run as a fix/upgrade backend alongside planemo’s lint/test; feed the Action that lints changed-tool PRs.	🔭 Potential: the most concrete “streamline reviews” path.
`gx-tool-db` / `galaxy_codex` (tool catalogs)	Feed check/upgrade signals at catalog scale for community-wide tool-health views.	🔭 Potential.
Galaxy core / ToolShed	Publish-time or load-time validation & repair.	🔭 Potential.
Repository-scale auto-fix	Walk a whole tool collection and apply the safe subset; gate a PR’s changed tools before merge.	🟡 Reference implementation: a bulk normalizer (Half A, `scripts/bulk_normalize.py`, proven on `tools-iuc`) and a forward-gate GitHub Action (Half B, `.github/actions/forward-gate/`) are built and shipped, but enforced on no repository (adoption is the maintainers’ call). The autonomous bot that opens the fix-PRs by itself remains 🔭 roadmap.
Beyond Galaxy (Dockstore, bio.tools, bioconda; CWL)	Tool-packaging & workflow ecosystems.	🔭 Roadmap.

See vs planemo for why the planemo relationship is complementary.

Source repositories (what the skill mines)

The repo-explainer skill draws real before/after examples and evidence from cloned tool repositories. It buckets them by leverage and never clones automatically; see the policy below.

Repo	Bucket	Suggested depth	Why
`tools-iuc`	tool corpus (primary)	deep (history)	the canonical IUC tools + PR history for real examples
`tools-devteam`	tool corpus	deep	Galaxy-team tools
`usegalaxy-tools`	tool corpus	deep	the usegalaxy.* common set
`galaxy-tool-repository-template`	authoring	deep	the canonical “new tool” starting point
bgruening, tools-galaxyp, RECETOX, tools-ecology, tools-metabolomics, tools-artbio, phac-nml	tool corpus (other communities)	deep	breadth of authoring styles (already in `corpus_sources.json`)
`planemo`, `planemo-ci-action`	integration	deep (small)	positioning + the CI-review path
`galaxy-language-server`	integration	deep (small)	the editor quick-fix path
`galaxy-mcp`	integration	deep (small)	agent-ecosystem positioning
`gx-tool-db`	catalog	deep (small)	catalog-scale signals
`galaxy` (core)	reference	shallow (reuse `.local/galaxy-src`)	source-of-truth for runtime behaviour; large
`galaxy_codex`	catalog	shallow	large catalog repo
`training-material`	education	never clone (~42 GB)	link, or sparse-checkout a tutorial path only if needed

Clone policy (the skill never clones for you)

The repo-explainer skill is side-effect-free by default. On each run it:

Inventories what’s present under .local/corpus/ and whether each clone is shallow.
Uses whatever is there to generate the guide.
For an absent or shallow high-leverage source, warns and prints the exact depth-tuned clone command, then asks: it does not pull bytes unprompted.
Flags the coverage gap in the output (e.g. “examples limited: planemo-ci-action not present. Clone to enrich”) rather than fabricating or silently narrowing.

Pulling a large repo with history is an explicit, human-consented action.

Note: the leverage manifest vs. corpus_sources.json

corpus_sources.json (workspace root) is the load-bearing seed list for the corpus sweeps and is consumed by scripts/corpus_check.py. The bucket/leverage/depth metadata above is kept here in the guide for now to avoid coupling the docs effort to that load-bearing file; a later change can fold purpose / leverage / depth fields into corpus_sources.json once the sweep tooling is updated to tolerate them.